Professor Masanori Shimono | Using AI to Reveal the Brain’s Hidden Patterns

In this study, AI models were used to predict brain activity. Data were obtained from in vitro mouse brain slices and recorded using electrode arrays that capture electrical signals from hundreds of neurons.

For example, the team applied a deep learning model to learn the firing patterns of neurons in one brain region and use those rules to predict activity in another. Interestingly, the model successfully predicted activity even in regions that were not physically connected, suggesting the brain follows universal principles in generating activity.

With further research advancements, the team has also succeeded in generating in vivo data, significantly improving the accuracy of their analytical methods. A paper summarizing these findings is expected to be published soon, marking an important milestone in this field.

One of the most noteworthy aspects of this research is its potential as an alternative to animal experiments. By using AI to predict brain activity, experiments that traditionally required live animals can now be conducted in a virtual environment, enabling more ethical and efficient research. This approach is particularly valuable for studying rare or hard-to-access data.

The findings of this study represent a significant step forward in both neuroscience and AI, paving the way for a future of science that is more ethical, efficient, and innovative.